Classifications

• • G—PHYSICS
  • G11—INFORMATION STORAGE
  • G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
  • G11B20/00—Signal processing not specific to the method of recording or reproducing; Circuits therefor
  • G11B20/10—Digital recording or reproducing
  • G11B20/18—Error detection or correction; Testing, e.g. of drop-outs
  • G11B20/1833—Error detection or correction; Testing, e.g. of drop-outs by adding special lists or symbols to the coded information

Definitions

• the invention relates to methods and apparatus for data reproduction, and in particular for the retrieval and decoding of data from a rotating carrier such as a Digital Versatile Disc (DVD).
• a rotating carrier such as a Digital Versatile Disc (DVD).
• DVD Digital Versatile Disc
• Mass-market optical data storage is familiar in the form of the well- known CD-ROM optical disc format.
• a historical review and technical description of these and other storage media can be found in The PC Technology Guide, itself available on CD-ROM, and at http://www.pctechguide.com/10dvd.htm. DVD expands upon the capabilities of CD-ROM, in terms of its nominal ("1x") data rate, as well as its data capacity.
• a DVD/CD decoder is described for example in WO-A-99/48097.
• Integrated circuits (ICs) incorporating suitable circuitry are available from Philips Semiconductors, for example product SAA7335.
• Multi-Beam One solution to the problem of delivering an ever higher data rate from a standard optical disc without increasing the spin speed is the so-called "Multi-Beam” approach.
• the read-back laser beam is split in an optical pick-up unit (OPU) into a number of separate beams focused on adjacent radial tracks.
• OPU optical pick-up unit
• the data can be read (in principle) at N times the rate, for a given speed of rotation of the disc.
• a multi- beam architecture for CD systems is described in WO-A-98/037555 (Zen Research). This description assumes that multi-track read out is achieved using a single laser combined with a diffraction grating which are arranged such that N neighbouring tracks are read in parallel. This means that the IC must have N data inputs and is capable of processing N data streams simultaneously. More elaborate arrangements with separate read heads are also known, in the context of CD readers, for example from US 5,465,244 (Kobayashi/Toshiba).
• the Multi-Beam approach also brings the possibility of reducing the power consumption of a drive if data rate is not an issue, as for a given data throughput, the disc spin speed can be proportionally reduced by a factor related to the number of beams.
• This approach gives significant power consumption benefits for portable equipment, as the power consumption of such drives is dominated by the power dissipated by the spindle motor and drivers.
• a DVD disc was designed to be read in a linear fashion, and although random access through jumps is part of the system design, the intention is that data is normally read as a stream, i.e. Data is continuous on the track from the start to the end of a file.
• All the data on a DVD (as is a CD) is arranged on a single spiral track running from the inner disc radius to the outer. If a multi- track approach is used it can be easily seen that the data being read by the individual pick-ups is effectively data from the same linear stream, but temporally shifted by one disc rotation, this means for an N beam system one disc rotation will yield N rotations worth of data.
• the data from the various pickups needs to be reassembled into one linear data set before passing on to the host application. After one rotation of the disc, the read-back head then needs to jump out N grooves to start acquiring the next consecutive block of data.
• the inventors have devised a range of architectures for a practical multi-beam DVD system (or a CD system), based on an understanding of the functions in the error correctors and by partitioning the memory into small, high-bandwidth blocks (suitable for integration) and larger, lower bandwidth blocks (allowing the use of standard commodity buffer RAM).
• These two types of memory blocks will be referred to herein as the "local” and “remote” memory, respectively.
• the "remote” memory will typically form part of the user data buffer within the DVD drive, but external to the decoder IC itself, but the invention is not limited to such an arrangement.
• the remote memory might be external to the chip, but separate from the user data buffer, or it might form part of a block of compact, slower memory on a part of the IC, separate from high performance buffers.
• the invention in a first aspect provides a method of reproducing data recorded in a first sequence, the data incorporating an error protection code applied on the basis of a predetermined size of data block, each error protection block comprising a series of several sub-blocks, the method comprising the following steps:
• step (e) Combining the remainder of said syndrome with the stored partial syndrome so as to obtain a complete syndrome for the block.
• the method may further comprise (f) Jumping to a new set of locations in said first sequence and reading in parallel a further group of N subsequences to cover a further portion of the first sequence, said further portion being contiguous or overlapping with the preceding one, while step (d) includes identifying said starting part of the block among the further group of sub-sequences.
• Said sub-blocks may be regarded as rows or groups of rows of a two- dimensional matrix, step (b) including performing an inner error correction on rows the matrix, while the syndrome accumulated in steps (c)-(e) is applied to columns of the matrix in a outer correction process.
• the method may be performed using distinct local and remote storage for temporary storage of syndromes, said partial syndrome being accumulated in local storage in step (b); the accumulated partial syndrome being transferred to said remote storage in step (c).
• Said transferring step (c) may be performed each time in association with said jumping step.
• the partial syndromes may be combined during reading of the starting part of the block, said partial syndrome being transferred from the remote storage back to the local storage in step (d) for accumulation of the remainder of said syndrome in said local storage.
• said partial syndrome and the accumulated remainder may be combined away from said local storage.
• the local storage may have space for approximately 2N syndromes.
• 3N-1 expensive syndrome buffers would be required to achieve the same processing without transferring the partial syndromes to cheaper storage during the rotation.
• the benefit of the invention can be seen in systems having fewer than 5N/2, or even fewer than 3N/2 syndrome buffers, depending on the memory bandwidth available.
• a system having 2N syndrome buffers is described below which combines the present invention with a syndrome double buffering technique.
• the data recorded in said first sequence may comprise data recorded in a spiral on a disc-like record carrier, such as a DVD.
• the invention in a second aspect provides a method of data retrieval in multi-track data read-out, the data comprising DVD data recorded on an optical disc, wherein partial error correction syndromes are calculated for parts of different ECC frames encountered by each beam, stored while other ECC frames are decoded and used as a basis for accumulating the complete syndrome.
• the partial syndromes are stored remotely from the syndrome generator, and restored when the remainder of the respective ECC frame is encountered by a different beam, for use in accumulating the complete syndrome.
• the invention further provides an apparatus for reproducing data stored on a record carrier, the apparatus comprising disc transport means including a pickup for reading data from the carrier, signal processing means for recovering data from the carrier and a decoder for decoding and error- correcting the data read from the carrier, in accordance with error correcting codes included therein, wherein the pickup is adapted for reading multiple- channels in parallel to recover plural sub-sequences of a first sequence of data recorded on the carrier, and wherein the decoder is arranged to implement a method according to the first or second aspect of the invention as set forth above.
• the invention yet further provides a decoder comprising input means for receiving in parallel plural sub-sequences of a first sequence of data to be decoded, and wherein the decoder is arranged to implement a method according to the first or second aspect of the invention as set forth above.
• the decoder may comprise an integrated circuit including inner and outer correction error correctors, and buffers for the storage of at least one outer error correction syndrome per channel, and means for transferring a partial syndrome to external memory after encountering the end of a block of data.
• Figure 1 illustrates schematically the structure of a known single-beam reproducing apparatus for DVD
• Figure 2 shows in block schematic form the process of decoding DVD data
• Figure 3 illustrates a DVD data sector format
• Figure 4 illustrates the structure of an ECC frame in DVD, including 16 data sectors and error correcting code words
• Figure 5 illustrates schematically the general structure of a multi-beam DVD reproducing apparatus
• Figure 6 illustrates the process of multi-beam read-out from an optical disc in DVD
• Figure 7 shows in block form a multi-beam DVD decoder circuit according to a first embodiment of the invention, with variations (a) and
• Figure 8 shows in more detail a row buffer and inner error corrector in the circuit of Figure 7;
• Figure 9 shows in more detail an outer error corrector in the circuit of
• FIG. 10 shows in more detail an external memory interface in the circuit of Figure 7;
• Figure 11 shows in block form a multi-beam DVD decoder circuit according to a second embodiment of the invention.
• Figure 12 shows in block form a multi-beam DVD decoder circuit according to a third embodiment of the invention.
• FIG. 1 shows the basic elements of an optical disc reproducing apparatus, in this case a DVD-ROM drive for use with a host computer, or part of a consumer DVD player for audio and video material.
• the key features are a transport mechanism 100 for the disc 102, an optical pick-up (OPU) 104 driven and focussed by various servo systems of known type, pre-amplifier 106, processing and buffering circuits 108, and an interface 110 to the computer, audio/video playback system, or other host.
• OPU optical pick-up
• the present disclosure concerns primarily the processing and buffering circuits 108, and the other elements will not be described in any more detail.
• Figure 2 shows that the read data path of a DVD-ROM system is a complex structure consisting of bit detection 200, EFM+ demodulation 202, error correction 204 & 206, storage buffer 208 and host interface 110.
• Error correction is performed in two stages: inner error correction 204 and outer correction 206. Both types of error correction assume that a syndrome (or "signature") is generated (212/216 respectively) for a particular codeword.
• a syndrome or "signature"
• Outer error correction is performed over the columns of the same sector using an outer codeword PO.
• the generated syndrome is then applied to an error correction module (214/218) that calculates the address and value of all correctable errors.
• the original data is recovered by applying the obtained corrections to the erroneous data.
• the outer correction process requires access to the buffer 208, which necessarily interrupts the flow of data arriving from the pick-up via the inner corrector.
• a FIFO buffer 220 is provided at the output of row buffer 210. Having applied the corrections to the data, it is transferred via host interface 110 to the user.
• the various elements 200-220 may be implemented by circuits dedicated to that function, or they may be implemented by a suitable combination of programmable hardware and software achieving the same function.
• FIG. 3 illustrates the DVD data sector format for data stored on disc 102.
• Raw computer data is broken into 2048-byte sectors, as in CD-ROM.
• Each DVD data sector is 2064 bytes long and comprises 12 bytes Identification (ID), which contains Physical Sector address & ID, the data, and four bytes of error correction and detection codes EDC. Note that this is shorter than a CD-ROM sector (2352 bytes).
• Figure 4 illustrates how 16 of the sectors shown in Figure 3 are combined into a Recording Block or error-correcting code (ECC) frame using a RS (Reed-Solomon) Product Code.
• ECC Recording Block or error-correcting code
• RS-PC is a block code, meaning that error correction (ERCO) operates over buffered blocks of data, these being the ECC frames.
• Each 2064-byte data sector occupies spread over 12 rows of the ECC frame, each row comprising 172 sector data bytes and 10 PI Bytes, which are the codewords (parity) added for the inner error correction.
• the sixteen PO bytes in each column form a 16-byte outer protection codeword for reach column of the ECC block.
• the sync patterns inserted every 91 bytes vary, according to their position in the block.
• the first sync pattern in each recording sector which is immediately followed by the ID field of the data sector, is unique within the ECC frame. In this way, the EFM+ decoder can readily identify the start of rows 0, 13, 26 ... and every 13th row within the ECC frame, where the sector ID field is to be found.
• DVD inner codewords correspond to a row of a DVD recording sector which consists of 172 data bytes plus 10 bytes parity check code. This implies that the row buffer 210 in Figure 2 is at least 182 bytes long. In practice, this row buffer is used to store one codeword during syndrome generation, one codeword being error corrected and acts as buffer for the input and output data. Therefore, it may be assumed that an upper bound of the buffer size is 4 times 182 bytes or 728 bytes. In addition to the data buffer, some storage for the syndromes is needed which, in case of inner correction, equates to 10 bytes.
• the DVD outer codewords correspond to the columns of the DVD recording sector and hence a much bigger amount of data must be considered. As there are 172 columns (length of a row) and 192 data rows in a recording sector, a buffer of 32 Kbytes size is required. Since it is expensive to integrate a buffer of this size into an IC, it is more efficient to calculate the syndromes while the data is being transferred into an external DRAM memory. This is possible as the correction process is only based on the syndrome and all corrections are applied in a read-modify-write operation to the external DRAM. This is indicated in Figure 2 with the two input channels to the buffer, one for the data after inner correction and the other for the results of the outer correction.
• Performing the error correction over the 172 columns of the recording sector implies that 172 syndromes of 16 bytes length must be stored, resulting in a storage requirement of 2752 bytes. Due to the fact that the data stream is still continuing while the calculated syndromes are applied to the error corrector, the FIFO buffer 220 is needed, with a typical capacity of 5 Kbytes (approximately 15% of the ECC frame).
• the error corrector typically has a gate count of well over 10,000 gates.
• the EFM+ demodulator (often realised as ROM lookup table) and syndrome generators are significantly smaller and each have a size less than 20% of the error corrector size. Therefore, the error corrector is often implemented as a shared resource in inner and outer error corrector.
• a multi-beam system might be envisaged, as is known already in CD-ROM systems such as that of WO-A- 98/37555 mentioned above.
• the entire data path of Figure 2 can be replicated N times for an N Beam system, the data from these N channels being recombined in a large RAM buffer.
• the chief disadvantage of this approach is that the error correction logic is replicated N times and the ERCO buffer (32kByte) and the Syndrome RAMS (2k7 byte) are replicated N times, which makes for a very expensive one-chip decoder.
• Figure 5 illustrates a more sophisticated multi-beam system, and one in which the syndrome double buffering and partial syndrome calculation mentioned in the introduction can be applied to advantage.
• a conventional (or high-spin speed) DVD disc transport 500 and disc 502 are supplemented with a multi-beam OPU 504 having N beams, and outputting read-out signals in N parallel channels.
• N 7 will be a typical figure in the embodiments which follow, although only four channels are shown in Figures 5 and 6, for clarity.
• Separate pre-amp stages 506a, b, c, d are provided, one per channel, and similarly channel processing up to and including inner ERCO processing is performed in parallel by circuits 508a, b, c, d.
• the main ERCO buffer can be placed into the external SDRAM 514, and parts of the error corrector may be shared between the channels.
• the extended functions of the buffer manager compared with that of a single- beam system, will be described later in relation to Figure 10.
• the inner corrector is conventional, and the outer syndrome generation (very high memory bandwidth) is on-chip, while the main buffers are off-chip in memory 514.
• a shared outer corrector 510 is used.
• Figure 6 illustrates the process of multi-beam read-out on an (exaggerated) spiral track of a DVD or CD.
• Channels a, b, c, and d are related to beam spots which follow four adjacent tracks on the disc.
• Beam a is the innermost of the four
• beam d is the outermost.
• the adjacent points can be regarded as four parallel tracks, it must be remembered that they are in fact simply points on one continuous spiral, so that, after one revolution, beam a reaches the position where beam b started, and so forth.
• ECC frames marked with an asterisk " * " in Figure 6 are only partially read by any one beam and this results in partial or out of order data arriving at the outer syndrome generators. This is not such a problem in CD-ROM systems, where the error correcting code is linear, and valid data can be decoded a short time after jumping to a new track location. In DVD however, decoding requires a complete ECC frame before valid data is obtained out of the inner and outer ERCO processes.
• sector IDs may be extracted from the inner corrector buffer, sixteen times per ECC frame, by which the data may be correctly located into the external buffer memory.
• the linear track length which must be passed before a sector ID is found is only around 4mm, so that the latency after jumping to a new track is only a small fraction of the rotation period.
• partial syndrome generation can be performed using the outer parity codewords (PO) which are interleaved with the user data in the recording sectors.
• syndrome generation essentially comprises multiplication of each received row by a predetermined polynomial function, the polynomial function ("alpha") being raised to a power each time, dependent on the position of the data within the block.
• alpha polynomial function
• the decoder thus preloads the syndrome alpha-power multipliers to be consistent with the position in the frame, which is known from the sector ID. This can be simply done from a lookup table, as only 16 start positions are possible (based on there being only 16 rows identifiable by sector ID field, in the ECC frame).
• Syndrome RAM approximately 2.7 Kbytes per ECC frame
• 16 Read- Modify-Write Cycles are necessary for each channel byte received.
• the apparatus disclosed herein the Syndrome buffers are allocated dynamically per channel (per beam) and buffers containing partial results are dumped into a cache area in the main stream buffer (memory 514 in Figure 5).
• the Cached data may be retrieved after one disc rotation when the start of the partially recovered ECC frame is again detected.
• the FIFO buffer between the inner and outer ERCO circuits is eliminated by use of the double-buffering of syndromes, which is explained more fully in our copending application mentioned above [498224]. Further expansion/description of the method is given below, in relation to several alternative embodiments.
• Multi-track DVD-ROM Systems As explained above, in order to reduce spin speeds while still maintaining or increasing performance and throughput, data is read from several parallel tracks in the proposed multi-track DVD-ROM systems. Different architectures can be envisaged, however, depending on the point in the data path ( Figure 2) at which the data that is received from the N parallel input channels is recombined into a single data stream. Obviously, the location of this recombination block has a significant impact on cost and performance of the resulting system. In a first detailed embodiment, recombination will be performed just after bit detection or EFM+ demodulation, resulting in a system where the main decoding electronics remain virtually unchanged from a single spot system.
• Figure 7(a) shows the block diagram of a system with buffer-less recombination where a multiplexer 720 is used to transfer the incoming data bytes into the EFM+ decoder.
• Alternative arrangement (b) of which the front end only is shown, has a separate EFM+ demodulator for each channel, and the multiplexer feeds into the row buffer of the inner error corrector section.
• the operation of the two variants embodiments is similar. However, the volume of data passing through the multiplexer will be much less in case (b).
• a dynamic syndrome buffer 722 is provided, in association with the outer ERCO circuitry, as will be described in more detail, with reference to Figure 9. As mentioned in the general discussion of the DVD data format, the
• EFM+ sync words which are inserted every 91 bytes allow the EFM+ decoders to identify the boundaries of rows and sectors in the ECC frame structure.
• the designers of the present apparatus have recognised that this valuable information is in fact available at an early stage in the data path. This allows data, particularly partial data, to be directed immediately to suitable buffer locations, rather than being held "in limbo" pending a deeper decoding of the sectors themselves.
• the various flags which are generated by the EFM+ demodulator to identify row and sector boundaries are shown as outputs of the multiplexer in Figure 7 (b): • data[31 :0], four data bytes at a time,
• N buffers of 182 bytes to accommodate incoming PI codewords
• N-1 buffers waiting for processing 1 buffer used during inner syndrome generation
• 1 buffer used in the inner error correction process 1 buffer for the data transfer to the outer error corrector.
• This is in total (2N+2) buffers.
• the row buffer must also contain logic that controls the input multiplexer, generates addresses for the physical buffer and identifies completed rows which are complete and ready for correction.
• control logic must ensure that corrections are written back to the correct buffer, fully corrected data is transferred to the outer error corrector and that unused buffer space is reallocated to the input buffering process.
• the buffer-less implementation of the stream recombination implies that outer parity codewords PO from N different recording sectors are presented to the outer corrector, thus resulting in the simultaneous generation of N syndromes. Therefore, additional control logic is needed to associate incoming data with a particular syndrome buffer and, when a particular syndrome has been completed, hand it over to the error corrector. Also, the control logic must ensure that the corrections are written to the correct address in the user data buffer.
• the syndrome buffer must store the N syndromes that are currently being calculated, N-1 syndromes that are awaiting corrections and 1 syndrome used in the current error correction process. Furthermore, it can be seen that before and after a jump only parts of an ECC sector are transferred via an input channel. This means that after a jump has been completed, up to N-1 partially calculated syndromes are calculated for ECC sectors whose first rows have been missed. The missing rows are read off the disc by a neighbouring input channel at the end of a disc rotation. The corresponding partial syndromes could be stored in the syndrome buffer 722. However, in order to minimise the size of this on-chip buffer, they are in the novel system stored in the off-chip user data buffer 724. If all syndromes were stored locally, 3N-1 syndrome buffers are needed, resulting in a total buffer size of 55040 bytes.
• the amount of syndrome buffers reduces to 2N. If also the N-1 syndromes awaiting correction are stored in external memory, there will be only N+1 buffers. In low performance systems, it may be possible to store N-1 of the syndromes in generation in external memory thus reducing the number of buffers to 2. Note that the described method of syndrome calculation requires that all incoming data bytes are pre-multiplied with the correct alpha value. Note also, that the size of the external user data buffer is determined by twice the amount of data on a disc rotation which is around 2 MByte. This is because the corrected data is placed into the buffer in a non-linear way and only after a full disc rotation it is possible to obtain a serial data stream.
• the data rate in each of the channels is about 18 MByte/sec which equates to an overall transfer rate over 120 MByte/sec.
• a bus width of 32-bit is suitable here.
• a clock frequency of 75 MHz is more than efficient to accept the data bytes from the individual channels, group them into a 32-bit word and multiplex them out of the stream recombination block.
• the central element in the inner correction section is the row buffer 710 as it is used to reconstruct the data coming in from EFM demodulation into Pl-codewords, identify completed codewords and initiate syndrome generation, hand syndromes over to the error corrector and ensure that corrections are written back into the row buffer. Furthermore, the data after inner correction must be transferred to the outer error correction section.
• the row buffer 710 is notionally divided, as shown by dotted boundaries in Figure 8, according to these functions. Of particular importance is the bandwidth of the row buffer. This is because four independent tasks read and write data out of the memory at a high rate. The combined data rate is determined by:
• two syndrome generators 712a and 712b work in parallel at an operating frequency of 75 MHz.
• This clock frequency is also a good choice for the error correction module which operates approximately twice as fast as the syndrome generator. Hence a single instance is sufficient.
• Control logic 800a and 800b is provided to control the operations of buffer arbitration, control of syndrome generators and error corrector, buffer assignment.
• the flags output by the multiplexer 721 in Figure 7 (b) are inputs to the control logic.
• Bandwidth is assigned in a fixed priority scheme where each of the tasks is serviced in a pre-defined order, for example a round-robin scheme.
• the four tasks refer to the diagram in Figure 8: 1. Store the incoming data into the row buffer such that incoming data quadlets are stored together with other data of the same PI codeword. To achieve this:
• the control logic monitors the data input and waits for the data_valid signal.
• the channel_nr is used to generate the task_select signal.
• the buffer offset is programmed by the control logic 800a at the start of a new PI codeword.
• a counter is incremented after a data item has been transferred.
• the control logic maintains a list of completed PI codewords together with the base address in the buffer.
• a buffer offset is programmed by the control logic while a running count is generated inside ACU 804.
• Outer Syndrome Handling and Error Correction Figure 9 shows the Outer Syndrome Handling and Error Correction 716, 718, 722 in more detail.
• syndrome generation for the PO codewords is a problem due to the high bandwidth requirements. For each data byte that is transferred into the outer error correction section, 16 syndrome bytes must be read, updated and written back into the memory. This problem is further complicated by the fact that four data bytes are received at a time, at a rate of 30 MWords/sec from the inner correction section.
• two syndrome generators 716a, 716b running at 75 MHz. Pipelined construction of the module allows a 16-byte syndrome and a data item to be processed every clock cycle. The two combined generators provide then a processing capability of 150 MByte/sec.
• the first issue in syndrome caching is the bandwidth required for the retrieval of completed syndromes out of the buffer. Assuming a clock frequency of 75 MHz and an access rate of 63 MHz, it is apparent that 85% of the bandwidth is used, 15% is still available. It can be shown that in situations where more than two ECC frames are completed at the same time, there is not sufficient time to copy a syndrome from the syndrome buffer into a different location: an available bandwidth of approximately 35% is needed.
• syndrome double buffering is used in addition to caching of partial syndromes.
• Double buffering of syndromes means that two syndrome buffers are used for each channel, one buffer being used for the generation of a syndrome set, the other containing completed syndromes before they are applied to the error corrector. Once a syndrome has been transferred to the error corrector, it is subsequently replaced with the error magnitude and location obtained from the error corrector.
• Syndrome double buffering is claimed in our co-pending application [ID 498224], mentioned above.
• a 16-byte syndrome For each incoming data byte, a 16-byte syndrome must be updated: • Each incoming data quadlet is stored in a latch, two data bytes are utilised in the first available clock cycle, the two remaining bytes in the following.
• Two syndromes are read, for example, out of banks B2 and B3.
• two syndrome generators update the syndromes assigned to the previous data byte while the results from the preceding cycle are stored in banks BO and B1.
• Error correction information is calculated from the syndromes:
• Control logic 900 generates an index signal that indicates to which ECC sector the corrections belong.
• the control logic monitors the IDs of incoming ECC sectors and once a sector is found that corresponds to a partial syndrome set, syndrome calculation is performed until it contains the information from the EDC frames that is not contained in the partial syndromes that is cached in external memory.
• the first partial syndrome is retrieved from external memory and combined with the second partial syndrome in the syndrome buffer. This requires a read-mod ify-write operation.
• Figure 10 shows in more detail the buffer manager (corresponding to block 513 in Figure 5) that receives the data and the outer corrections coming from the DVD error corrector ( Figure 9) and stores them in the external DRAM
• Interface logic 922 manages the addressing and transfer of input data being read from the disc, corrections, partial syndromes and output data being retrieved and streamed out via the host interface (512 in Figure 5).
• ACUs address calculation units
• the association of an incoming data item to a particular ACU is done via an index signal that is provided by the outer error correction section ( Figure 9).
• Address generation inside an ACU is be based on a running index, related to the number of incoming bytes, a part that is related to the EDC ID number, and a part that is programmable by the control CPU. In this way, it can be arranged that the data between two jumps is retrieved and sorted into the external memory without further CPU interaction.
• the apparatus monitors whether all EDC frames that are present on the corresponding disc tracks have been retrieved without error and stored in the DRAM.
• the buffer manager in the present embodiment has the following specific functionality (alternative arrangements are of course possible), provided by control logic 926.
• Each EDC frame is represented by a single bit (or by a set of bits) in a small register file. These bits indicate the status of each EDC frame including: not received, in transfer, data and corrections received, unrecoverable error.
• Jumping is initiated by interrupt to a controlling CPU, which may be on the same IC as the decoder. Similarly, an interrupt should be generated when some EDC frames are erroneous. In this case, of course, the jump should be delayed such that the EDC frames in question are read again on the following disc revolution, this time by a different input channel.
• the buffer manager requires also an additional channel with ACUs 928 for the caching of partial syndromes. This channel is used only for burst transfers and should have higher priority than other channels. This is because the corresponding data must be read from and written to the outer syndrome buffer which has only limited bandwidth available.
• the bandwidth requirements to the external memory is determined in the described 64x system by the rate at which data is retrieved from the channel side, the number of outer corrections and the host interface bandwidth.
• the data rate from the error corrector is 120 MByte/sec. This should, on average, also be the host interface data rate.
• a bandwidth of about 3.5 MByte/sec is required for the caching of partial syndromes (assuming a 11.5x spin speed, around 110 jumps are performed per second, and after each jump 6 partial syndromes of 16 times 172 bytes must be read and written).
• Row buffer control logic that is capable of generating addresses for the N input data streams and is able to identify when a row has been completed and syndrome generation as well as error correction must be initiated.
• Control logic in the outer error correction section that maintains a set of syndrome buffers and assigns those to the input channels. This logic also needs to identify which syndromes are complete and available for error correction. Furthermore, addresses or address indices for the user data buffer (data and corrections) must be generated.
• Buffer manager needs additional input that is used in the address generation for user data and corrections. It must be possible that the control logic in the outer error corrector section generates an address or address index that allows to associate the current user data or corrections with the memory segment that is allocated to a particular EDC or ECC frame. Additional address calculation units (ACUs) may be needed, depending on system implementation.
• ACUs address calculation units
• FIG. 11 shows an alternative multi-track read path based on EDC frames.
• Outer syndrome calculation could be performed in a system with N separate row buffers, N inner ERCO syndrome generators and N inner ERCO syndrome buffers.
• For the inner error correction a common error correction circuit is used dynamically in turn as inner syndromes become available.
• the outer error corrector in this example is shared between channels, having access to the syndromes that are complete, and waiting in the syndrome buffers, to be processed.
• the association of syndrome buffers to input channels and the handling of partially generated syndromes is dynamic, rather than this is implemented in an arbitration and control logic 950.
• the two syndrome buffers 952a and 952b in each channel are used in the following way:
• syndrome generation is stopped (also data transfer into user buffer memory can be stopped).
• the second syndrome buffer is available for syndrome generation in the normal way.
• Power consumption can be controlled by keeping clock frequency in the N channels to a minimum.
• the DVD decode path consists of bit detection, inner error correction, outer error correction and user data buffer.
• the first example has considered stream recombination after the bit detection/EFM+ decoding, while the second example recombined the data streams only after outer error correction.
• stream recombination occurs after inner error correction.
• Figure 12 shows a block diagram of this architecture.
• the skilled reader will by now understand the functions of the various blocks in the diagram.
• the inner corrections which are just based on distinct rows of the ECC frame or block are performed in N individual channels.
• the inner error corrector can be shared across the N channels.
• the resulting data (not just the syndromes) can be stored in a buffer 960 where the ECC frames are reconstructed.
• the data could be recombined using a multiplexer instead of a buffer, as used in the first example ( Figure 7).
• Figure 7 The implications of this on the syndrome handling in the outer error corrector section have already been described in detail, with reference to Figure 10 in particular.
• this structure which is a mixture of the architectures shown in Figure 7 and Figure 11 , does not provide any particular benefits or shortcomings.
• the most important shortcomings are probably that the cost of the inner correction sections are quite high while, at the same time, the sequential correction in the outer error correction section limits the performance.
• the system of Figure 12 requires arbitration circuitry for the inner error corrector module. Also it requires either • Buffer for raw data; this buffer should be integrated with the user data buffer in order to keep the pin count low. Control logic is needed that generates addresses for the data transferred via the N input channels, identifies whether all data from an entire disc revolution has been received, and reconstruction of the serial data stream. Also an additional input channel must be added to the user data buffer. OR
• Control logic in the outer error correction section that maintains a set of syndrome buffers and assigns those to the input channels. This logic also needs to identify which syndromes are complete and available for error correction. Furthermore, addresses or address indices for the user data buffer (data and corrections) must be generated.
• the actual bit detection, demodulation, syndrome and error correction as well as the host interface can be performed by existing blocks.
• DVD system nor to optical discs in general.
• other systems may be known or envisaged which employ a block-based error protection scheme, and where parallel processing of different parts of the data leads to the generation of partial syndromes and corrections.

Landscapes

• Engineering & Computer Science (AREA)
• Signal Processing (AREA)
• Signal Processing For Digital Recording And Reproducing (AREA)
• Error Detection And Correction (AREA)
• Optical Recording Or Reproduction (AREA)

Applications Claiming Priority (3)

Publications (1)

Family

ID=9905729

Family Applications (1)

Country Status (7)

Families Citing this family (23)

Family Cites Families (10)

• 2000
  • 2000-12-22 GB GBGB0031439.3A patent/GB0031439D0/en not_active Ceased
• 2001
  • 2001-12-10 JP JP2002553770A patent/JP2004517429A/ja not_active Withdrawn
  • 2001-12-10 EP EP01272135A patent/EP1346360A1/en not_active Withdrawn
  • 2001-12-10 WO PCT/IB2001/002416 patent/WO2002052559A1/en not_active Application Discontinuation
  • 2001-12-10 CN CN01808467A patent/CN1426583A/zh active Pending
  • 2001-12-18 US US10/023,338 patent/US6877126B2/en not_active Expired - Lifetime
  • 2001-12-20 TW TW090131669A patent/TW556161B/zh not_active IP Right Cessation

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events